Date: Sat, 18 Mar 2006 01:07:16 -0800 From: soralx@cydem.org To: freebsd-hardware@freebsd.org Subject: Re: FreeBSD shutting down unexpectedly Message-ID: <200603180107.16268.soralx@cydem.org> In-Reply-To: <1142613543.28676@swaggi.com> References: <1142613543.28676@swaggi.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> > be sure to check Tcpu in BIOS before trusting mbmon (on my board, mbmon > > show CPU temperature ~20*C lower than BIOS; chipset temp & other data > > seem to be OK) > > > > see Intel's datasheets for power consumption, Tmax, etc. > > you can calculate approximate operating temperature by multiplying heat > > dissipation value (which is about equivalent to power consumption) by the > > cooler's thermal resistance (expressed in K/W or *C/W). > > You were right, I checked the BIOS hardware monitor and found out that > mbmon was reporting the temperature 20*C lower then the BIOS. This brings > my avg Tcpu to around 55*C which seems somewhat high considering Intel's > datasheet says my processor is rated for a max of 69*C. > > Should I be worried? well, 1.26Ghz Socket370 PIII has Pmax of ~29.5W (which can be considered rather low these days). If it's cooler has performance of 0.3K/W (not unreasonable for a 1U server CPU cooler), this would give (29.5 W)*(0.3 K/W)=8.85K rise, which at SATP translates to heatsink's temperature of ~34*C. Inside the case the temperature is higher (you need to measure it), but 40*C is rather common, so we get a cooler that's ~50*C hot. The interface between CPU's core and cooler's surface has limited thermal conductance; this could raise processor's Tcore by a couple degrees. Mainboards' temperature probes are usually quite inaccurate, and 5*C error is very well possible. So, as you can see, 55*C is not all too bad. However, if you measure the temperature with the case open and CPU idle, and cooler's performance is same or better than assumed, you'd better not rely on this processor. In fact, 55*C is somewhat too high in any case, considering that there exists additional heat dissipation path through mainboard. I'd check the thermal interface btw CPU and cooler first. Is the heatsink sitting level on the core? Is there a nice thin layer of clean thermal compound between them? Fan turning at good RPM? Then I'd check Vdd with a scope (or at least a DMM). Is it at the right level and clean? At this point I would think twice before replacing the CPU. Overheating it could have created some kind permanent latchups (shorts from Vdd to Vss directly), which would result in higher power consumption, but this isn't likely, plus you'd definitely see some instability or erros in CPU operation. So I personally don't think that CPU damaged by overheating can consume more power, but be stable, and then suddenly die some day; correct me if I'm wrong. It is not very likely that you CPU was damaged by overheating too. It might not have been stable when overheated (no kidding!), but I belive the mainboard should power it down before it reaches temperature at which permanent damage results. Timestamp: 0x441BB5DA [SorAlx] http://cydem.org.ua/ ridin' VN1500-B2
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200603180107.16268.soralx>